K404
06-20-2007, 09:11 AM
http://img522.imageshack.us/img522/8347/boxes1wr5.jpg (http://img522.imageshack.us/my.php?image=boxes1wr5.jpg)
http://img507.imageshack.us/img507/1544/cpuvrmpe5.jpg (http://img507.imageshack.us/my.php?image=cpuvrmpe5.jpg)
The board:
Aux Molex plug is to supply additional power to PCI-E slots. Placement is pretty poor- its beside the 24-pin ATX plug!
6 S-ATA ports are controlled by ICH9R chipset, with an additional 2 controlled by a Jmicron chip. (Called “Gigabyte RAID”)
http://img522.imageshack.us/img522/7043/hdtacht166ich9rfz8.jpg (http://img522.imageshack.us/my.php?image=hdtacht166ich9rfz8.jpg)
compared to what I got on the P965/ ICH8R:
http://img103.imageshack.us/img103/7245/samsunghd160hjt166singlov9.jpg (http://img103.imageshack.us/my.php?image=samsunghd160hjt166singlov9.jpg)
6-phase CPU power. Gigabytes top-end board the DQ6, has 12 power phases. This should not matter much at all. The components used on both boards are the same, but in general, it’s the quality of the phases that counts, not the quantity. Intel spec states that the minimum phase count is 4.
The CPU socket center is very sp****ly populated. Only 4 caps present here, out of a possible 18. Not sure what to think of this. It shouldn’t affect stability much, but time will tell. On the plus side, the CPU area is good for extreme cooling solutions. Some air coolers might have narrow clearance margins due to the shape of the NB heatpipe solution, but I cannot comment as I don’t use air cooling for Intel CPUs. Ever.
The NB and SB heatsink assemblies are bolted to the board, instead of the usual plastic clips or hooks. Nice touch. The rear of the NB and SB areas (on the back of the board) have small heatsinks as well, which may or may not help cooling, but hinder heatsink changes slightly. The VRM sinks revert to sprung plastic clips though.
The 8-pin CPU power plug is located between 2 section of the heatpipe assembly. It shouldn’t be too hard to reach or work with though. Four of the 8 pins are masked off to make sure users with a 4-pin CPU plug don’t use the wrong side.
Sound comes from a Realtek RTL 889A chip. Nothing spectacular, but I’m hoping this sounds better than the recent Asus offerings, which produced too much crackle for my liking.
Generally the board layout is good. Gigabyte have already had time with this general layout when designing the P965 boards. A few tweaks here and there have been made, but the ground-work already done helps explain why these boards have been so fast to market.
Maximum Available Voltages:
CPU: 2.35V (yes...>>2V!)
MCH: +0.375V
FSB: +0.35V
PCI-E: +0.35V
DDR2: +1.55V (from 1.80V standard) (yes... 3.35V is an option!)
Primary running notes: I installed windows with the CPU at 400x8. We can thus mark 400FSB off as easy. The 2nd thing to strike home is how cool...no -COLD- this chipset appears to run. Some might assume this means the heatpipes are doing their job. This is not the answer- the NB/SB combo spit out so little heat that the heatpipe solution is absolute overkill. Once initial testing is done I will make sure heatsink contact is good.
some Multimeter probing on the board soon zeroed in on the voltage readpoints. Setting 2.10 volts for the RAM (+0.3V)gives a real voltage of 2.165V so the general tracking looks pretty average.
http://img261.imageshack.us/img261/7917/sbsataic2.jpg (http://img261.imageshack.us/my.php?image=sbsataic2.jpg)
I`m still working on read-points. Should have most of the important ones done within a couple hours, then onto FSB testing :)
The audio quality is much much cleaner than the recent Asus solutions. Bass does not distort and the bottom end of music seems a bit fuller.
Update: VMCH seems a bit hit n miss with each BIOS step, adds 0.05V if +0.15V is added, and adds 0.25V if +0.25V is requested
450FSB needs an MCH and FSB voltage bump. Still fine-tuning ideas- I know the board has a load more in it :)
EDIT 2:
480FSB seems fine, 500 doesnt want to play. I suspect my E67 is the wall here, and will put me around 492FSB but i`m not done yet.
Droop and VCore: 1.45V BIOS gives 1.452V windows idle and 1.451V SuperPi. 100% load gives: 1.451V. Voltage readpoint was taken from one of the missing CPU Inductor points.
SuperPi: still working on combos at 3600MHz. Best 1M so far at 400x9 is 14.062 secs, not tweaked.
Memset values dont match all the BIOS value, and vice-versa. In F4 BIOS, I cant see tRFC or refresh-refresh delay options! They naturally default to slack numbers depending on RAM multiplier/RAM MHz so I know theres a performance hit there.
Having some trouble getting to the VDIMM pins for a direct measurement m3kk, but i`ll work at it.
EDIT 3:
Swapped chips- 500FSB is a piece of cake but I suspect this CPU hits a wall too, just after 500FSB. Will try FSB again next time I have Dry Ice.
Updated BIOS to F5d (from F4) Not noticing any big differences.
http://img507.imageshack.us/img507/1544/cpuvrmpe5.jpg (http://img507.imageshack.us/my.php?image=cpuvrmpe5.jpg)
The board:
Aux Molex plug is to supply additional power to PCI-E slots. Placement is pretty poor- its beside the 24-pin ATX plug!
6 S-ATA ports are controlled by ICH9R chipset, with an additional 2 controlled by a Jmicron chip. (Called “Gigabyte RAID”)
http://img522.imageshack.us/img522/7043/hdtacht166ich9rfz8.jpg (http://img522.imageshack.us/my.php?image=hdtacht166ich9rfz8.jpg)
compared to what I got on the P965/ ICH8R:
http://img103.imageshack.us/img103/7245/samsunghd160hjt166singlov9.jpg (http://img103.imageshack.us/my.php?image=samsunghd160hjt166singlov9.jpg)
6-phase CPU power. Gigabytes top-end board the DQ6, has 12 power phases. This should not matter much at all. The components used on both boards are the same, but in general, it’s the quality of the phases that counts, not the quantity. Intel spec states that the minimum phase count is 4.
The CPU socket center is very sp****ly populated. Only 4 caps present here, out of a possible 18. Not sure what to think of this. It shouldn’t affect stability much, but time will tell. On the plus side, the CPU area is good for extreme cooling solutions. Some air coolers might have narrow clearance margins due to the shape of the NB heatpipe solution, but I cannot comment as I don’t use air cooling for Intel CPUs. Ever.
The NB and SB heatsink assemblies are bolted to the board, instead of the usual plastic clips or hooks. Nice touch. The rear of the NB and SB areas (on the back of the board) have small heatsinks as well, which may or may not help cooling, but hinder heatsink changes slightly. The VRM sinks revert to sprung plastic clips though.
The 8-pin CPU power plug is located between 2 section of the heatpipe assembly. It shouldn’t be too hard to reach or work with though. Four of the 8 pins are masked off to make sure users with a 4-pin CPU plug don’t use the wrong side.
Sound comes from a Realtek RTL 889A chip. Nothing spectacular, but I’m hoping this sounds better than the recent Asus offerings, which produced too much crackle for my liking.
Generally the board layout is good. Gigabyte have already had time with this general layout when designing the P965 boards. A few tweaks here and there have been made, but the ground-work already done helps explain why these boards have been so fast to market.
Maximum Available Voltages:
CPU: 2.35V (yes...>>2V!)
MCH: +0.375V
FSB: +0.35V
PCI-E: +0.35V
DDR2: +1.55V (from 1.80V standard) (yes... 3.35V is an option!)
Primary running notes: I installed windows with the CPU at 400x8. We can thus mark 400FSB off as easy. The 2nd thing to strike home is how cool...no -COLD- this chipset appears to run. Some might assume this means the heatpipes are doing their job. This is not the answer- the NB/SB combo spit out so little heat that the heatpipe solution is absolute overkill. Once initial testing is done I will make sure heatsink contact is good.
some Multimeter probing on the board soon zeroed in on the voltage readpoints. Setting 2.10 volts for the RAM (+0.3V)gives a real voltage of 2.165V so the general tracking looks pretty average.
http://img261.imageshack.us/img261/7917/sbsataic2.jpg (http://img261.imageshack.us/my.php?image=sbsataic2.jpg)
I`m still working on read-points. Should have most of the important ones done within a couple hours, then onto FSB testing :)
The audio quality is much much cleaner than the recent Asus solutions. Bass does not distort and the bottom end of music seems a bit fuller.
Update: VMCH seems a bit hit n miss with each BIOS step, adds 0.05V if +0.15V is added, and adds 0.25V if +0.25V is requested
450FSB needs an MCH and FSB voltage bump. Still fine-tuning ideas- I know the board has a load more in it :)
EDIT 2:
480FSB seems fine, 500 doesnt want to play. I suspect my E67 is the wall here, and will put me around 492FSB but i`m not done yet.
Droop and VCore: 1.45V BIOS gives 1.452V windows idle and 1.451V SuperPi. 100% load gives: 1.451V. Voltage readpoint was taken from one of the missing CPU Inductor points.
SuperPi: still working on combos at 3600MHz. Best 1M so far at 400x9 is 14.062 secs, not tweaked.
Memset values dont match all the BIOS value, and vice-versa. In F4 BIOS, I cant see tRFC or refresh-refresh delay options! They naturally default to slack numbers depending on RAM multiplier/RAM MHz so I know theres a performance hit there.
Having some trouble getting to the VDIMM pins for a direct measurement m3kk, but i`ll work at it.
EDIT 3:
Swapped chips- 500FSB is a piece of cake but I suspect this CPU hits a wall too, just after 500FSB. Will try FSB again next time I have Dry Ice.
Updated BIOS to F5d (from F4) Not noticing any big differences.